Supporting Online Materials
Bayesian reconstruction of the whole-genome phylogenetic tree of S.
pneumoniae.
We have tested the reliability of the phylogenetic reconstruction shown in Fig. 1
using a Bayesian Markov Chain Monte Carlo method [69]. The obtained tree is
shown in Fig. S1. The structure of the tree is similar to the structure of the tree in
Fig. 1. The clades conserved between the two trees are indicated as lineage I-VI
in Fig.1. The major differences are the positon of Group VI, which in this
reconstruction appears to be closely related to clade I, and the position of the
serotype 7F strain CDC1087, which does not belong to Group I, but instead
belongs to a monophyletic brnch with the serotype 5 strain 70585. Using the
possibility to define test statistics in the program BEAST [69], we have tested the
probability
that the six lineages I-VI are monophyletic. We obtain 100%
probability for groups II-VI. For group I, the probability is 0 if we include also the
strain CDC1087, and 100% if we only include the serotype 3 strains), suggesting
that while the composition of lineages II-VI is robust, the inclusion of CDC1087
in lineage I depends on the tree reconstruction method.
Whole genome phylogenetic analysis generalizes MLST-like typing
methods
To quantify the degree of correlation between phylogeny and dispensable
genome, we measured the association between the lineages of Fig.1 and
presence/absence of dispensable genes present in more than one strain (see also
Table 1). For each of these genes, we quantified the strength of the association
between the allelic variant and the lineages of Fig.1 by computing the association
index V (see Methods). A value of V=1 means perfect association, while V=0
indicates random association between allelic variants and lineages. The average
association value was 0.5 (median of the P-values equal to 0.07) indicating, in
most cases, a non-random association between presence/absence of dispensable
genes and lineages. Repeating the analysis for the association of the dispensable
genome with the CCs we found an average value of V=0.82 (median of the Pvalues equal to 0.007).
Given the extent of homologous recombination, we investigated the extent to
which the allelic variants of core genes correlated with lineages I-VI inferred by
molecular phylogenetics. We identified 680 core genes that were present in
single copy in each genome, and classified the distinct alleles of each of these
genes. The average association is 0.747 (median of the P-values equal to
3.9 106 ) indicating a strong association between alleles and lineages. Not
surprisingly since CCs are always contained in a single lineage, their association

was even stronger (V equal to 0.94, median of the P-values equal to 5.3 1011 ).
We conclude that, knowing the CC of a strain, we can predict the allelic variant of
most core genes with high level of confidence.

Taken together, these results indicate that there is a strong and statistically
significant association between the allelic form of core genes and the
classification into lineages I-VI and CCs. A significant association exists also
between presence/absence of dispensable genes and both lineages and CCs. This
association is weaker than in the case of allelic variants of core genes, probably
due to the more dynamical nature of the dispensable genome, which is rich in
mobile elements [3].
Clustering based on presence/absence of dispensable genes. The pattern of
presence/absence can be used to obtain a hierarchical clustering, that identifies
groups of related strains. In Fig.S8 we show a hierarchical clustering based on
presence-absence of dispensable regions; red boxes mark the cluster exceeding
the 98% confidence level using a multiscale bootstrap resampling method [70,
71].
The clusters obtained with this method coincide with monophyletic
branches of the phylogenetic trees shown in Fig.1, confirming the correlation
between genomic content and genetic distance. Compared to Fig. 1, there are
variations in the mutual position of some groups. For instance, Group V
containing all serotype 1 strains is split by this analysis into tree distinct clusters,
one corresponding to the CC306 strains, the second formed by the CC90 strains
670_6B and SP18 whose dispensable genome is more closely related to the
Taiwan19F_14 and G54 strains, and the third corresponding to the
CC217/CC2296 strains, that are more strictly related to the serotype 3 strains.
Interestingly this is confirmed by an analysis of the repeats present in these
genomes, because NCTC7465, P1031 and P1041 share some repeat families that
are only found in the serotype 3 isolates in this dataset (see below). Group I,
containing all serotype 3 strains and the serotype 7F CDC1087 strain, has been
split in a group containing only the serotype 3 strains, while the strain CDC1087
is not strictly related to any other strain, suggesting that, as indicated also by the
Bayesian posterior probability, the relative order of the A45 and CDC1087
branches in Fig.1 is probably an artifact of the phylogeny reconstruction method.
Insertion Sequences are more variable than other phylogenetic markers.
In total, intact representatives of 19 different insertion sequence (IS) types could
be identified in the dataset. This includes ISSpn7, found to have caused an
insertion disrupting the hyaluronate lyase gene of S. pneumoniae strain SP18,
and ISSpn8. The frequency distribution of IS elements is variable. For instance,
we found that IS200 is present as a single copy (either intact or degenerate) in
each genome, and there were no observed instances of transposition. In contrast,
other elements, such as ISSpn5, were absent from some isolates, but increased
the copy number in others. It seems likely that this dissemination within
genomes occurs relatively quickly compared to other changes in the genome, e.g.
alterations in the sequence of the MLST loci. For example, IS1515 in the serotype
1 genome Sp033038 and Sp032672 contain 16 and 14 copies of this transposon
respectively, whilst strain Sp061370 (of the same ST) has only one copy,
suggesting that loss or gain of IS1515 has occurred since these strains have
diverged from each other. However, there is little phylogenetic signal in the
distribution of repeats, which is likely to be a consequence of the high turnover
of such elements: the majority of CDS annotated as being transposons appear to
be non-functional, suggesting they are remnants of past IS expansions. Hence the
IS complement of each pneumococcal chromosome appears to be in a state of
flux, with new families imported and expanding to replace those that are eroded
by inactivating mutations.
Fig.S1. Bayesian phylogenetic tree obtained using the SNPs of the whole genome
alignment of the 44 genomes of S.pneumoniae.
Fig. S2 Hierarchical clustering of the stains based on the presence/absence of
dispensable regions. Red numbers indicate the percentage p-value of the branch
computed using the Approximate Umbiased (AU) method (see Methods), while
green numbers indicate the bootstrap support. Blue and red bars indicate MLST
Clonal Complexes and Sequence Types, respectively.
Table S1. Sequenced strains used in this study. For each strain, we report name,
genome length, number of predicted coding sequences, locus of isolation, disease
or carriage status, sequencing status, source of the sequence, accession number,
sequencing technology and average sequence coverage. For pneumococcal
strains, also ST and CC determined by the eBURST algorithm [21] are reported.
*)Finished genomes are complete genomes that were inspected for proper
assembly and manually inspected for ambiguous consensus base calls.**)These
Whole Genome Shotgun projects have been deposited at DDBJ/EMBL/GenBank
under the accession xxxx00000000. The version described in this paper is the
first version, xxxx01000000
Table S2. For each strain (rows) we report the number of genes not conserved in
the other sequenced genomes (columns).
Table S3. Distribution of the genes from Table 3 which are not shared by all S.
pneumoniae strains. Green squares indicate presence, gray squares indicate
absence of the gene.